Akash Ganapathi

CEO,

[Akash Ganapathi @Akash_UG]

@yacineMTB 1 way doors and many categories of investing are obvious exceptions… risk is a thing. But yea, experimentation can teach you more than just theorizing… too balanced a take for X but idgaf about engagement

[Akash Ganapathi @Akash_UG]

Lots of virtue signaling about practicing gratitude, not enough about practicing forgiveness. It’s a lot more meaningful if you forgive others before you’ve done something wrong yourself.

[Akash Ganapathi @Akash_UG]

@DearS_o_n Great point. Sounds like a decision any man can make, idiot.

[Akash Ganapathi @Akash_UG]

@tryopine @SocketSecurity Nice.

[Akash Ganapathi @Akash_UG]

Watching wannabe “coaches” try to build personal brands by presenting second hand insights and -EV hot takes

[Akash Ganapathi @Akash_UG]

@stats_feed Rage bait. But if helpful, this study has been widely criticized for using outdated and regionally “extrapolated” data (they just guesstimated regions from neighboring countries), culturally biased tests, and generally flawed methods. It also ignores socioeconomic factors.

[Opine @tryopine]

Your CRM is great—if you’re closing deals. But for presales? It’s like bringing a spork to a steakhouse.

Read more: https://t.co/75f1ywRw7Q

[Opine @tryopine]

Our CEO, @Akash_UG used to rely on Homerun for his presales team. But like many teams, he ran into the same pain points: endless data entry, continued work in excel, and an outdated process that was more frustrating than efficient. So, we built https://t.co/sGbNPk0M2n

[Akash Ganapathi @Akash_UG]

@hamptonism The key to the problem is how do you stop multiple from running simultaneously? They don’t know you have one bullet, but they can see you’re just one guard. They’ll think they have a chance working together. You use them against each other. If any of gang1 escapes I’ll kill gang2

[Akash Ganapathi @Akash_UG]

@shl Assuming you mean the memetic usage: “they let the intrusive thoughts win” — it comes from the vestiges of our child-like curiosity, like when a kid wants to touch a stove even though their parents said not to. Part of how we learn is retesting/confirming.

[Akash Ganapathi @Akash_UG]

@Codie_Sanchez Creating copies of your originality as a value prop is an oxymoron - humanity’s creative value is overrated. Just a fancy automation experience.

[Akash Ganapathi @Akash_UG]

@AustinKelleher Clean work man.

[Andrew Ng @AndrewYNg]

This week, Google announced a doubling of Gemini Pro 1.5's input context window from 1 million to 2 million tokens, and OpenAI released GPT-4o, which generates tokens 2x faster and 50% cheaper than GPT-4 Turbo and natively accepts and generates multimodal tokens. I view these developments as the latest in an 18-month trend. Given the improvements we've seen, best practices for developers have changed as well.

Since the launch of ChatGPT in Nov 2022, with key milestones that include the releases of GPT-4, Gemini 1.5 Pro, Claude 3 Opus, and Llama 3-70B, many model providers have improved their capabilities in two important ways: (i) reasoning, which allows LLMs to think through complex concepts and and follow complex instructions; and (ii) longer input context windows.

The reasoning capability of GPT-4 and other advanced models makes them quite good at interpreting complex prompts with detailed instructions. Many people are used to dashing off a quick, 1- to 2-sentence query to an LLM. In contrast, when building applications, I see sophisticated teams frequently writing prompts that might be 1 to 2 pages long (my teams call them “mega-prompts”) that provide complex instructions to specify in detail how we’d like an LLM to perform a task. I still see teams not going far enough in terms of writing detailed instructions. For an example of a moderately lengthy prompt, take a look at Claude 3’s system prompt. It’s detailed and gives clear guidance on how Claude should behave.

This is a very different style of prompting than we typically use with LLMs’ web user interfaces, where we might dash off a quick query and, if the response is unsatisfactory, clarify what we want through repeated conversational turns with the chatbot.

Further, the increasing length of input context windows has added another technique to the developer’s toolkit. GPT-3 kicked off a lot of research on few-shot in-context learning. For example, if you’re using an LLM for text classification, you might give a handful — say 1 to 5 examples — of text snippets and their class labels, so that it can use those examples to generalize to additional texts. However, with longer input context windows — GPT-4o accepts 128,000 input tokens, Claude 3 Opus 200,000 tokens, and Gemini 1.5 Pro 1 million tokens (2 million just announced in a limited preview) — LLMs aren’t limited to a handful of examples. With many-shot learning, developers can give dozens, even hundreds of examples in the prompt, and this works better than few-shot learning.

When building complex workflows, I see developers getting good results with this process:
- Write quick, simple prompts and see how it does.
- Based on where the output falls short, flesh out the prompt iteratively. This often leads to a longer, more detailed, prompt, perhaps even a mega-prompt.
- If that’s still insufficient, consider few-shot or many-shot learning (if applicable) or, less frequently, fine-tuning.
- If that still doesn’t yield the results you need, break down the task into subtasks and apply an agentic workflow.

I hope a process like this will help you build applications more easily. If you’re interested in taking a deeper dive into prompting strategies, I recommend Microsoft's Medprompt paper (Nori et al., 2023), which lays out a complex set of prompting strategies that can lead to very good results.

[Original text (with links): https://t.co/UOtLDza1Vh ]

[Akash Ganapathi @Akash_UG]

Google’s phone ads are the fucking worst

[Nicholas A. Christakis @NAChristakis]

The University of Chicago, as usual, striking the right balance. Worth reading in full. It is exactly right.

President Alivisatos’ Note on the Encampment (April 29, 2024)

Dear Members of the University Community, Just a few hours ago, a group of students established an encampment on the Main Quad as a form of protest. This particular tactic is now in widespread use at universities across the country. At some, encampments have been forcibly removed, with police arresting students and faculty in chaotic scenes that are disturbing. At others, encampments have persisted, despite attempts to shut them down with force. In some cases, encampments have resulted in major disruptions to learning and the activities of the university community.

Free expression is the core animating value of the University of Chicago, so it is critical that we be clear about how I and my administration think about the issue of encampments, how the actions we take in response will follow directly from our principles, and specific considerations that will influence our judgments and actions.

The general principle we will abide by is to provide the greatest leeway possible for free expression, even expression of viewpoints that some find deeply offensive. We only will intervene when what might have been an exercise of free expression blocks the learning or expression of others or that substantially disrupts the functioning or safety of the University. These are our principles. They are clear.

Two recent examples illustrate how we bring these principles into real action. First, last quarter a student group secured university permission to cover a large fraction of the Main Quad with a massive Palestinian flag consisting of thousands of tiny colored flags. The exhibit was accompanied by signage exhorting passersby to “Honor the Martyrs,” and it was staffed by students at tables during certain hours. Those students could explain to passersby why they thought it important to feature this installation, why they thought that language was appropriate, and any other views occasioned by their installation. While this protest and accompanying message were offensive to many, there was no question that it was an exercise of free expression. It stood for weeks until the end of the approved time, at which point the student group removed it, making way for others to express their views in that space as they might see fit. This example should make it clear that we approach the issue with no discrimination against the viewpoints of those participating in this encampment. We adhere to viewpoint neutrality rigorously.

As a second illustrative example, in November, a group of students and faculty undertook an occupation of Rosenwald Hall, a classroom and administration building. That was a clear disruption of the learning of others and of the normal functioning of the University. After repeated warnings, the protesters were arrested and released. They are subject now to the University’s disciplinary process, which is still pending. In short, when expression becomes disruption, we act decisively to protect the learning environment of students and the functioning of the University against genuinely disruptive protesters.

There are almost an unlimited number of ways in which students or other members of the University community can protest that violate no policies of the University at all; the spectrum of ways to express a viewpoint and seek to persuade others is vast. But establishing an encampment clearly violates policies against building structures on campus without prior approval and against overnight sleeping on campus.

I believe the protesters should also consider that an encampment, with all the etymological connections of the word to military origins, is a way of using force of a kind rather than reason to persuade others. For a short period of time, however, the impact of a modest encampment does not differ so much from a conventional rally or march. Given the importance of the expressive rights of our students, we may allow an encampment to remain for a short time despite the obvious violations of policy—but those violating university policy should expect to face disciplinary consequences.

The impact of an encampment depends on the degree to which it disrupts study, scholarship, and free movement around campus. To be clear, we will not tolerate violence or harassment directed at individuals or groups. And, disruption becomes greater the longer the encampment persists. With a 24-hour presence, day after day, we must for example divert police resources away from public safety for our campus and our community.

If necessary, we will act to preserve the essential functioning of the campus against the accumulated effects of these disruptions. I ask the students who have established this encampment to instead embrace the multitude of other tools at their disposal. Seek to persuade others of your viewpoint with methods that do not violate policies or disrupt the functioning of the University and the safety of others.

Sincerely,

Paul

[Akash Ganapathi @Akash_UG]

@itspatmorgan One of the more serious resources for B2B SaaS https://t.co/ygwWHOTscP
Always available as a resource to you man if you want my opinion!

[Akash Ganapathi @Akash_UG]

@nikitabier If your apps can’t go viral yet make your tweets go viral instead…

[Akash Ganapathi @Akash_UG]

Don’t forget, all of your accomplishments are worthless in the eyes of a screaming baby